Turn your AI security position into something a board can act on in ten minutes: a one-page brief with a headline, a risk heat map, what you are doing and the decisions you need, plus a 5-slide outline with speaker notes. Risks come from the ISO/IEC 27090 threat themes and are written in plain words, not jargon.
Free: the brief, the outline, Markdown export and print to PDF. Pro: the full deck with appendix. Your draft stays in this browser. Not sure of your maturity? Estimate it with the ISO 27090 Maturity Timeline (Pro).
No download, no card
The sample is filled in so you can see the output. Replace it with your own, or start blank.
1. The meeting
2. AI systems in use
Name them the way the board would recognise them.
3. Top risks
Drawn from the ISO/IEC 27090 threat themes, shown in board language. Rate how likely and how bad, and add one line on why it matters here.
Our AI can be tricked by hidden instructions
Prompt injection (direct and indirect)
High
Confidential data could leak through AI
Sensitive data leakage
High
We rely on AI we did not build
Supply chain and model provenance
Medium
4. Current maturity
The same five steps as the ISO 27090 Maturity Timeline (Pro), which estimates how long each step takes. It is a planning scale, not an external rating.
Aware: There is an owner and a first list of AI systems. Protection is patchy.
5. Incidents and near misses
6. What we are doing
Short actions in plain words.
7. Decisions and budget
Board of directors · Q4 2026
AI security: Example Co
We use AI in 3 systems and 2 risks need attention now. We are at "Aware" and plan to reach "Defined" by end of Q2 2027. We need 1 decision from you today.
Risk heat
SevereHidden instructionsSeriousAI suppliersData leaking via AIMinorLess likelyPossibleLikely
Top risks
1. Our AI can be tricked by hidden instructions(High) The customer assistant reads web content and can open support tickets.
2. Confidential data could leak through AI(High) Staff use of public AI tools is not yet controlled.
3. We rely on AI we did not build(Medium) Two AI suppliers have not completed our security review.
What we are doing
Now: Aware → target Defined by end of Q2 2027.
Done: Publish an AI acceptable-use rule for staff (October 2026)
Under way: Complete the list of every AI system and supplier (December 2026)
Planned: Test the customer assistant against planted instructions (January 2027)
Incidents this period: 1. A member of staff pasted a customer contract into a public AI tool.
Decisions needed
Approve one company-wide AI assistant and block unapproved public tools · we recommend: Approve · by this meeting
Maturity steps are a planning scale, not an external rating. ISO/IEC 27090 is guidance and cannot be certified.
5-slide outline with speaker notes
Slide 1
Where we stand on AI security
We use AI in 3 systems and 2 risks need attention now. We are at "Aware" and plan to reach "Defined" by end of Q2 2027. We need 1 decision from you today.
AI in use: Customer service assistant; Staff AI assistant; Fraud scoring model.
Where we are: Aware. There is an owner and a first list of AI systems. Protection is patchy.
Speaker notes: Open with the headline and stop. The board wants three things: is AI being used safely, what could go wrong, and what do we need from them. We use AI in 3 systems, some of them provided by suppliers, which means part of the protection sits with those suppliers. Our current position is "Aware" on a five-step scale we use to plan, not an external rating.
Slide 2
The risks that matter
High: Our AI can be tricked by hidden instructions. The customer assistant reads web content and can open support tickets.
High: Confidential data could leak through AI. Staff use of public AI tools is not yet controlled.
Medium: We rely on AI we did not build. Two AI suppliers have not completed our security review.
Speaker notes: Walk through the top risks in order. For each: what it means for us in one sentence, then how likely and how bad. "Our AI can be tricked by hidden instructions" is rated medium likelihood and high impact: Someone plants text in an email, document or web page and our AI follows it instead of us. "Confidential data could leak through AI" is rated high likelihood and medium impact: Staff paste sensitive data into AI tools, or an AI answer exposes data to the wrong person. "We rely on AI we did not build" is rated medium likelihood and medium impact: A supplier's model, data or software could be flawed or compromised, and we inherit the problem.
Slide 3
What has happened, and what we learned
September 2026: A member of staff pasted a customer contract into a public AI tool. Lesson: We need an approved tool and clear rules on what may be shared.
Speaker notes: Be direct about what happened and what it cost. A member of staff pasted a customer contract into a public AI tool. Impact: No known harm; the contract was not confidential beyond commercial terms.. What we changed: We need an approved tool and clear rules on what may be shared. A near miss reported is a sign the reporting works.
Slide 4
What we are doing about it
Done: Publish an AI acceptable-use rule for staff (October 2026)
Under way: Complete the list of every AI system and supplier (December 2026)
Planned: Test the customer assistant against planted instructions (January 2027)
Goal: reach "Defined" by end of Q2 2027. We have a policy, a full AI inventory and agreed protections for the riskiest systems.
Speaker notes: 1 of 3 actions are done and 1 are under way. Link each action back to a risk on the previous slide. Moving from "Aware" to "Defined" means: We have a policy, a full AI inventory and agreed protections for the riskiest systems.
Slide 5
What we need from you
Decide: Approve one company-wide AI assistant and block unapproved public tools (we recommend: Approve), by this meeting
Speaker notes: Ask for each decision explicitly and wait for an answer. Close by naming when you will report back.
Want the numbers the board should see every quarter? Practise choosing them in the Board Metrics Lab (Pro).
TL;DR
What this pack does
4 points+
A board briefing on AI security should fit on one page: a headline, the risks that matter, what you are doing, and the decisions you need.
Enter the AI systems in use, the top risks (picked from the ISO/IEC 27090 threat themes and shown in plain words), current maturity, incidents or near misses, decisions needed and any budget ask.
You get a one-page brief with a risk heat map and a 5-slide outline with speaker notes. Export to Markdown or print to PDF free.
Pro adds a full deck: a standalone, printable slide file with an appendix of systems, every rated risk, the action plan, the incident log and likely board questions.
What boards ask about AI security
Directors rarely ask how an attack works. They ask whether the organisation knows where it uses AI, who is accountable, what could go wrong and how badly, whether suppliers are covered, and what it will cost. A good briefing answers those questions in that order and ends with a clear ask, so the meeting produces a decision rather than a discussion.
Keep it to the four or five risks that matter most and say which ones you are not worried about. Report near misses: they show the reporting works. Use the same structure every quarter so the board can see movement. For the measures to track over time, practise in the Board Metrics Lab (Pro).
Lead with one sentence that says where you stand, then show no more than four or five risks in plain words with how likely and how bad each is, what you are doing about them, and exactly what you need from the board. Keep technical terms for the appendix. Boards decide on risk appetite, accountability and money, so end with those asks.
What should be in an AI security board report?+
Which AI systems the organisation uses and who supplies them, the top AI risks in business terms, current maturity and the target, incidents and near misses, progress on actions, and the decisions or funding needed. This tool builds each of those sections from your inputs.
Why does it avoid technical terms like prompt injection?+
Directors need to understand the consequence, not the technique. Each ISO/IEC 27090 threat theme is shown as a plain sentence, for example "Our AI can be tricked by hidden instructions" for prompt injection. The technical name and its public reference appear only in the Pro appendix, for anyone who asks.
What is free and what is Pro?+
The inputs, the one-page brief, the risk heat map, the 5-slide outline with speaker notes, the Markdown export and print or save as PDF are free. The full deck, a standalone slide file with a title slide, the five slides with notes and an appendix, is part of PlayCISO Pro.
Is the maturity level an ISO/IEC 27090 rating?+
No. ISO/IEC 27090 is guidance, defines no maturity levels and cannot be certified. The five steps are PlayCISO's planning scale, the same as in the ISO 27090 Maturity Timeline. Present it to the board as your own assessment, not an external score.
Where is my data stored?+
Only in this browser's local storage, so your draft is there when you come back on the same device. Nothing is sent to PlayCISO.